Evaluating Drug Interaction Databases: Results from an Exploratory Analysis

Katherine Saunders, PharmD
Ambulatory Oncology Clinical Pharmacy Specialist
AU Health
Augusta, GA

Patients are taking an increasing number of medications, including those used to treat cancer and those required for treating comorbidities. Patients and healthcare team members rely on oncology pharmacists to be experts on all these medications. An exciting aspect of oncology practice is the number of new oral antineoplastic (OA) agents coming to the market and the potential for new approvals for existing OAs. Given that many of these drugs may come to market under expedited review through the U.S. Food and Drug Administration (FDA), approval may be granted without a formal drug-drug interaction analysis. With the introduction of these drugs into practice comes a challenge for oncology pharmacists: how to best screen a patient’s medications for clinically relevant drug interactions when initiating an OA agent. 

Dr. John B. Bossaer and Dr. Christan M. Thomas aimed to shed light on this clinical question through an exploratory evaluation of existing drug interaction databases, a common resource  for clinicians. This research, titled “Drug Interaction Database Sensitivity with Oral Antineoplastics: An Exploratory Analysis,” was published in March 2017 in the Journal of Oncology Practice.¹ Although these databases have been previously evaluated in regard to their sensitivity in detecting drug interactions, those studies did not include OAs. The research team selected 20 drug interactions encountered in actual clinical practice and deemed to be clinically relevant. Clinically relevant was defined as a drug interaction that would require a change in therapy, such as selection of a new drug, a change in dose, or more frequent monitoring. This definition is more liberal than those used in the past; however, the authors felt that it more accurately reflected real-world practice. These 20 interactions include a variety of types, as well as newer agents that lack formal drug interaction studies. Examples include the interaction between pazopanib and amiodarone, which results in an increased risk of QTc prolongation, and the interaction between idelalisib and rivaroxaban, which results in an increased risk of bleeding. Well-known interactions, such as that between capecitabine and warfarin, were also included. These 20 interactions were then processed through five electronic databases commonly encountered in clinical practice: four of these were designed for the healthcare professional (Epocrates, Facts & Comparisons, Lexi-Interact, and MicroMedex), and one was oriented toward patients and non–healthcare providers (Drugs.com). The distribution of sensitivity across groups was assessed using the Cochran Q test.

The authors report that the sensitivities for each database varied. MicroMedex was found to have a sensitivity of 70% (95% confidence interval [CI], .46–.87), Facts & Comparisons, 70% (CI, .46–.87), Epocrates, 90% (CI, .69–.98), Lexi-Interact, 95% (CI, .73–.99), and Drugs.com, 95% (CI, .73–.99). The difference in their detection of clinically significant drug interactions was statistically significant, with a pvalue of .016. In addition, classification of interactions (major, moderate, not detected, etc.) varied among the databases. Only three interactions (those between idelalisib and phenytoin, bosutinib and voriconazole, and ibrutinib and voriconazole) were classified the same across all five databases. Even if the interactions labeled as major or moderate were considered equivalent, the databases had an agreement rate of 45% (n= 9).

The authors state that these results are not intended to indicate the superiority of one database over another. On the contrary, the variability among the databases, both in detecting an interaction and classifying it, suggests to the researchers that clinicians should use at least two electronic databases when checking a drug interaction. Databases that are transparent in the analysis of a drug interaction (for instance, including links to data supporting the interaction) are preferred. The authors also discuss the need for obtaining a detailed medication history from the patient during clinical encounters rather than relying on a medication list reported in the electronic medical record. 

The authors also discuss areas for future research and initiatives and highlight the need for standardization in analysis of drug interactions. The authors note the lack of criteria for defining drug interactions with OAs, which likely contributes to the variability seen in these databases. Furthermore, this study omitted interactions not deemed clinically significant and therefore does not assess the specificity of these databases. Future endeavors that focus on specificity may address alert fatigue, a concern related to the extensive use of electronic resources by those working in the healthcare system. Theoretical drug interactions should also be included in these analyses because they may not come to light until several years after the drugs’ introduction into clinical practice.

This exploratory study shows the growing complexity of oral antineoplastics. Clinicians are faced with new agents for which formal drug-drug interaction studies may be lacking. In addition, patients are taking an increasing number of medications for other chronic diseases, and these medications are often prescribed by other providers. Given the high cost and insurance restrictions associated with OAs, patients may be using specialty pharmacies in addition to their preferred local pharmacy, making it difficult for pharmacists to conduct an accurate drug utilization review. Finally, although these databases have the potential to identify many of the clinically relevant drug interactions, the decision about what action to take is patient-specific and is driven by many factors. The work of Dr. Bossaer and Dr. Thomas examining the usefulness of these drug databases in detecting drug-drug interactions, as well as the potential shortcomings of these databases, highlights the need for the clinical expertise of an oncology pharmacist as a member of a patient’s cancer care team.  

Reference

1. Bossaer JB, Thomas CM. Drug interaction database sensitivity with oral antineoplastics: an exploratory analysis. J Oncol Pract. 2017;13(3):217-222.